boincmgr.exe --namehost or /n does not work on Windows #2739
Comments
|
Was not able to reproduce this local debug version (7.14.1) and remote boinc client running on RasperryPi: |
|
Reproducable on 7.14.0. I did not realize that there was a new version. It still does not work. I get no error messages, is there any way to enable a log to get it to show what the manager is doing? |
|
What OS do you have inside VM? |
|
I've reproduced this across a LAN (Windows --> Windows). No error messages found in either the BOINC data directory or AppData. |
|
Ok, in Debug mode it works fine. I'll try to reproduce this in Release mode later this evening |
|
Correction - I was testing across the LAN, with the BOINC client already running on the local machine. I used the /nd flag, described by --help as
to prevent clashes. Testing again without /md, but - in both cases - supplying the GUI RPC password, it connected as expected. Maybe the behaviour of /nd should be considered. In remote mode, the Manager can't start the client anyway: maybe its apparent separate role in "don't connect to running client" should be documented. |
|
It makes no sense to use /nd flag because it affects local client only. If address and password are specified boincmgr should not try to start local client. This behavior should be verified also. |
|
Adding /nd has no effect on outcome. I do not have a GUI password set, but for giggles I manually set a simple 5 letter password, restarted the boinc-client inside the VM and reconnected manually with password and success as expected. When using the /p or --password= on a command line it does not connect and as previously reported there is no log to view on my end. Shouldn't there be to diagnose issues such as these? Edit: I just tested from another Windows 10 machine and I get the same outcome. Linux boinc_mgr is working fine regardless of the destination OS the client is running on; Win 10 or Linux all success |
|
I tried next scenario with 7.12.1 on both machines: @RichardHaselgrove, could you please share your steps to reproduce this? Thanks |
|
Ok, seemed to have narrowed down something. If I install to the default suggested directory "\Program Files\Boinc" it WILL connect automatically. |
|
I was trying with two Windows 7 machines, both on same LAN with local 192.168.x.y addressing. Using your numbering, Environment 1 is downstairs, logged in, BOINC started from Manager, Manager minimised, monitor off. Environment 2 is my daily monitoring machine, runs BoincView (so I can see the status of Env1 at all times), also runs BOINC client autostarted by Manager. Both machines have the same GUI RPC password set - which may become relevant later. For testing, I exited the manager on Env2, leaving the client running. I opened a command window at the BOINC programs directory, and tested various permutations of
Consistently, NO connection was made to the remote (Env1) machine when (and only when) /nd was on the command line. Connection was made with /n hostname /p password (as it should), and also - to my surprise - with just /n hostname. That should have caused an error, but maybe the use of the same password on both machines avoided that - the running client or registry entry for the data directory may have supplied the necessary access to the local password. Both machines are set to display the exit dialog. Manager windows with remote connections in general do not show the exit dialog, but in this test there was one exception: with /n hostname /nd, the exit dialog appeared when exiting the Manager for the next test. |
That sounds like your program location. Did you test any changes to the data directory location?
That sounds as if the second W10 machine isn't authorised as a 'remote manager' in the remote_hosts.cfg file for the clients you're trying to reach. |
|
It is authorized because if I manually connect it works fine. The /n flag is the hangup, I can always connect manually via File -> Select Computer If I use the localclient (host machine) from that same C:\Archives\Boinc (and ..\Data) it will start the local client and connect fine. The issues may be that my Data directory falls under the actual Boinc directory, but I dont understand why this is an issue (coding issue beyond my skill level) Default suggested location: I take that back, it does not work |
|
@RichardHaselgrove, pr #2745 should fix your case. Please, review it |
|
Ok, tried to reproduce this. |
|
Just tried a brand new install of Win10 build 1607. Boinc 7.14.2 Installed Boinc to default directory. Tried /n 192.168.2.21 and --namehost=192.168.2.21. It does not connect There must be something going on with W10 if you're having success on W7. |
|
Same results for Win10 (BOINC Manager 7.12.1 used)
|
|
So found a little headway here. I reinstalled W10 a 3rd time. I did NOT reinstall Boinc post W10 reinstall but used the previous install directory. I ran boincmgr.exe from \boinc_old It connects!!! No matter what my destination was, either a VM on the same machine or another on the local network, boingmgr.exe /n 192.168.2.x /p password connected perfectly. So I saw the new 7.14.2 devlopment build and installed it. Now I CANNOT connect via /n no matter the install location; it does not connect via command line Uninstalling Boinc does not resolve it. No matter what even running boincmgr.exe /n from the previous old directory no long works. So there is some sort of setting I'm guessing in the registry that is now set. Update: If I go through the registry and delete ALL registry items that have anything to do with Boinc, I can now connect via command line perfectly. So the install of boinc sets something in the registry that will cause it not to work. Edit 2: Found the offending registry setting
So it looks like boincmgr.exe cannot connect to a remote host when it is set to not load the local client, however a local client is NOT spawned if using the /N flag. Only the manager is. What caused it to fail was unchecking "run the client" in Options -> Other options. This sets the RunDaemon key in registry to 0 which then causes the manager to not connect properly with the /N switch. |
|
Ok, this case is already handled by #2745 |
|
Should be fixed in #2745 |
I have not been able to get the --namehost or /n flags to work on my computer running a VM and trying to connect to the boinc running inside the VM
Connecting through the Manager gui "File -> Select Computer" works fine
if I attempt to connect to the VM from my Linux Mint machine using the correct syntax for linux:
boincmgr -n 192.168.2.22, it does connect properly
Neither the current version 7.12.1 nor the previous 7.10.2 work
my command line is pretty simple: C:\Archives\Boinc\boincmgr.exe /n 192.168.2.22 C:\Archives\Boinc\boincmgr.exe --namehost=192.168.2.22 does not work either
I have used the --verbose flag and it does not seem to work either as none of the err files are showing anything unless they are meant for the client not the manager, but I get no errors at all on the screen using either command-line. However if I type the flag wrong, it will give me a popup and tell me the argument
The text was updated successfully, but these errors were encountered: